As the demand for mobile data traffic continues to increase, regulatory bodies have proposed to allow radar and communication systems to co-exist in the same frequency band. The success of co-existence would depend on how well the interference which one system exerts to the other can be controlled. Recently proposed co-existence approaches use a precoder at the communication and/or radar system, with the precoding matrices optimally designed to control the interference to the radar. However, this could potentially pose a privacy risk for the radar, as the precoding matrix assigned to the communication system contains implicit information about the radar.  In this talk we consider the scenario in which an adversary has obtained the precoding scheme of a smartphone co-existing with a radar, and based on that, study the possibility of the adversary inferring sensitive radar information, for example, the radar location. We consider an adversary inference attack that follows a machine learning approach. The adversary divides the search region into cells. The adversary separately trains a classifier for each cell, with training data consisting of smartphone respective precoder matrices. Then, for every grid cell, the classifier makes a binary prediction as to whether the radar is or is not present in the cell.